Differences

This shows you the differences between two versions of the page.

--- tech:slurm [2020/06/22 17:50] – [CUDA NVIDIA TESLA Infos] kohofer
+++ tech:slurm [2021/03/19 12:14] – kohofer
@@ Line 267: / Line 267: @@
 ===== Compute Nodes =====
 A compute node is a machine which will receive jobs to execute, sent from the Controller, it runs the slurmd service.
@@ Line 342: / Line 341: @@
+===== Modify user accounts =====
+Add user
+  sacctmgr add user <usernme> Account=gpu-users Partition=gpu
+Modify user, give 12000 minutes/200 hours for usage
+  sacctmgr modify user misegata set GrpTRESMin=cpu=12000,gres/gpu=12000
+Restart the services:
+  systemctl restart slurmctld.service
+  systemctl restart slurmdbd.service
+Check status:
+  systemctl status slurmctld.service
+  systemctl status slurmdbd.service
 ===== Links =====
+[[https://slurm.schedmd.com/slurm_ug_2011/Basic_Configuration_Usage.pdf|Basic Configuration and Usage]]
 [[https://slurm.schedmd.com/overview.html|Slurm Workload Manager Overview]]
@@ Line 362: / Line 385: @@
 ====== Modules ======
+Add different python versions using spack!
+. First see which python versions are available:
+  spack versions python
+  ==> Safe versions (already checksummed):
+.8.2  3.7.7  3.7.4  3.7.1  3.6.7  3.6.4  3.6.1  3.5.2  3.4.10  3.2.6   2.7.17  2.7.14  2.7.11  2.7.8
+.8.1  3.7.6  3.7.3  3.7.0  3.6.6  3.6.3  3.6.0  3.5.1  3.4.3   3.1.5   2.7.16  2.7.13  2.7.10
+.8.0  3.7.5  3.7.2  3.6.8  3.6.5  3.6.2  3.5.7  3.5.0  3.3.6   2.7.18  2.7.15  2.7.12  2.7.9
+==> Remote versions (not yet checksummed):
+.10.0a6  3.8.7rc1  3.7.6rc1   3.6.8rc1   3.5.7rc1   3.4.9     3.4.0     3.1.2rc1   2.7.9rc1  2.6.6     2.4.5
+.10.0a5  3.8.7 ....
+  .....
+. now select the python version you would like to install:
+  spack install python@3.8.2
+. Now you need to add a modules file
 ===== Python =====
@@ Line 391: / Line 437: @@
 ==== Create modules file ====
+**PYTHON**
   cd /opt/modules/modulefiles/
@@ Line 407: / Line 454: @@
 </code>
+**CUDA**
+  vi /opt/modules/modulefiles/cuda-10.2
+<code>
+#%Module1.0
+proc ModulesHelp { } {
+global dotversion
+puts stderr "\tcuda-10.2"
+}
+module-whatis "cuda-10.2"
+set     prefix  /usr/local/cuda-10.2
+setenv          CUDA_HOME       $prefix
+prepend-path    PATH            $prefix/bin
+prepend-path    LD_LIBRARY_PATH $prefix/lib64
+</code>
 ===== GCC =====